Detecting Spiking Queries

1. A computer-implemented method of identifying a legitimate search query spike using a computing system having memory, processor, and data storage subsystems, the computer-implemented method comprising: receiving a plurality of search query requests from one or more user input devices; identifying one or more spikes in the received search query requests; clustering the identified spikes together according to a temporal or textual correlation; determining a rate of acceleration in which each spike in the search query requests is received via the processor of the computing system; comparing the determined rate of acceleration for the clustered identified spikes with a similar temporal behavior of stored clusters; identifying a particular clustered spike of received search query requests as a malicious attack when the determined rate of acceleration exceeds a first threshold level and a comparison to temporal behavior is lower than a second threshold level; and storing non-malicious clustered spikes of received search query requests and results as one or more content groups in the data storage subsystem of the computing system for comparison and query suggestions to future related search query requests.

2. The computer-implemented method of claim 1 , wherein the identifying one or more spikes comprises: determining an instantaneous velocity of each query request from the query stream for a particular parameter, calculated as an inverse of a difference in time between an instant query request and a most recent query request for the particular parameter; calculating an updated weighted average velocity from a combination of a previous weighted average velocity and a weighted instantaneous velocity for the particular parameter; calculating an instantaneous acceleration of each query request for the particular parameter as a difference between the updated weighted average velocity and the previous weighted average velocity, per the difference in time between an instant query request and a most recent query request; and determining an updated weighted average acceleration from a combination of a previous weighted average acceleration and a weighted instantaneous acceleration for the particular parameter.

3. The computer-implemented method of claim 2 , wherein a spiking occurs when the weighted average velocity is greater than a base velocity just prior to an acceleration occurrence and the weighted average acceleration is equal to or greater than a percentage of a peak acceleration value at every instant in which acceleration is occurring.

4. The computer-implemented method of claim 2 , wherein a first contribution by the previous weighted average velocity can be varied with a second contribution by the instantaneous velocity.

5. The computer-implemented method of claim 2 , wherein the previous weighted average velocity and the instantaneous velocity identify false spiking by averaging out instantaneous spikes.

6. The computer-implemented method of claim 1 , wherein the clustering further comprises clustering the identified spikes together with a plurality of similar stored search query results.

7. The computer-implemented method of claim 1 , wherein the clustering produces a reduced number of false spikes, improves classification accuracy for detecting popular queries, and detects seasonal queries by comparing clustering across a number of time periods.

8. A computer-implemented method of producing popular search query results using a computing system having memory, processor, and data storage subsystems, the computer-implemented method comprising: receiving a search query request from a user input device; identifying a spike in a query stream comprising the received search query request and other incoming search query requests; temporally correlating the spike in the query stream with relevant content from a plurality of historical indices as a result of searching said historical indices; correlating the spike in the query stream with relevant content from a plurality of fresh indices as a result of searching said fresh indices, wherein the fresh indices contain information and results from recently crawled content sources; determining a rate of acceleration in which the spike in the query stream is received via the processor of the computing system; comparing the determined rate of acceleration with a temporal behavior of similar stored search queries; analyzing results from searching the historical indices and the fresh indices to determine if the search query request should be clustered with an existing group of search query results via a grouped content algorithm; identifying a particular clustered spike in the query stream as a malicious attack when a rate of acceleration exceeds a first threshold level and a comparison to temporal behavior is lower than a second threshold level; storing non-malicious clustered spikes of the query stream as one or more seasonal galleries in the data storage subsystem of the computing system; prioritizing results of the search query request according to an age and size of identified clustered results; and communicating the one or more seasonal galleries and the prioritized results of the search query request to a user output device.

9. The computer-implemented method of claim 8 , wherein the non-malicious clustered spikes comprise an acceleration rate which is below a peak acceleration rate and an average velocity which has increased a certain percentage over the base velocity.

10. The computer-implemented method of claim 8 , wherein the fresh indices are refreshed more frequently and the content sources are crawled more frequently when a spike is detected.

11. The computer-implemented method of claim 8 , wherein the one or more seasonal galleries comprise a calendar of clustered seasonal results retrieved from the data storage subsystem.

12. The computer-implemented method of claim 11 , wherein one or more of the stored seasonal galleries are combined with one or more similar query stream requests to provide query suggestions to the user input device.

13. A computer-implemented method of identifying and clustering queries that are increasing in popularity using a computing system having memory, processor, and data storage subsystems, the computer-implemented method comprising: receiving a search query request from a user input device; identifying a spike in incoming query stream activity comprising the search query request; temporally correlating the spike in the incoming query stream activity with relevant content from a plurality of historical indices as a result of searching said historical indices; correlating the spike in the incoming query stream activity with relevant content from a plurality of fresh indices as a result of searching said fresh indices, wherein the fresh indices contain information and results from recently crawled content sources; analyzing results from searching the historical indices and the fresh indices to determine if the search query request should be clustered with an existing group of search query results; prioritizing results of the search query request according to an age and size of identified cyclic clustered results; and communicating the prioritized results of the search query request to a user output device.

14. The computer-implemented method of claim 13 , wherein the searching a plurality of historical indices comprises extracting information from previously stored cyclic clustered results with similar characteristics to the search query request.

15. The computer-implemented method of claim 13 , further comprising communicating query suggestions from said analyzing.

16. The computer-implemented method of claim 13 , further comprising suggesting temporally relevant online advertisements to the user device.

17. The computer-implemented method of claim 13 , wherein the fresh indices are continually updated with information from relevant classified queries in reaction to a popularity of a query.

18. The computer-implemented method of claim 13 , further comprising: identifying a particular clustered spike of received search query requests as a malicious attack when a determined rate of acceleration exceeds a first threshold level and a comparison to temporal behavior is lower than a second threshold level.

19. The computer-implemented method of claim 13 , further comprising: identifying legitimate queries that are increasing in popularity by determining a weighted average query velocity from a weighted existing query velocity and a new instantaneous query velocity.

20. The computer-implemented method of claim 19 , wherein one or more of the weighted average query velocity and the weighted existing query velocity are modified according to a desired number of past sample points.